How has COVID impacted Domestic violence rates in Chicago?
Which communities have been the most impacted by this?
How did rates during the quarantine compare to rates when we weren't in lockdown?
We used data from the City of Chicago Crime Data to review reported domestic violence crimes, Chicago Health Atlas to make any comparisons and correlations with domestic violence rates and neighborhoods, and Google Trends API (PyTrends Library) to assess domestic-violence related search terms.
Our focus was narrowed to the timeframe of strict lockdown under the Restore Illinois plan (phase 1), which lasted from March 20th, 2020-May 28th, 2020. This is a total of 69 days. We then selected the 69 days before lockdown and 69 days after lockdown to ensure no statistical effects were seen due to uneven time periods.
Upon review of the City of Chicago crime data statistics and Chicago Health Atlas, we decided to look for alternate data sources in order to verify if help for domestic violence was being sought outside of the law enforcement/criminal justice system. We used the Google Trends API to search four key terms related to domestic violence to see if there was a significant increase in search terms during lockdown compared to pre- or post-lockdown.
We narrowed down the crime data to primary categories within the larger domestic category to exclude crimes that appeared to be outside of the scope of this question.
A limitation of the Google Trends API is that the data are sampled from the larger amount of searches per day, leading to slight variability from API call to API call. The trends also only encompassed the entire State of Illinois for interests over time, because this was the smallest region available for our needs.
The Chicago Health Atlas Hardship Index draws from six factors: crowded housing, poverty, education, unemployment, dependency, and income.
This heatmap shows the weighted number of domestic violence crimes per rounded latitude and longitude during 2020.
This figure examines the number of crimes per the top four category during the three time periods (pre-, during, and post-lockdown) to see if primary categories and counts varied.
Overall domestic violence crime seems to be highly concentrated in certain communities in the south and west sides of Chicago.
High concentrations of communities of color are in the south and west sides of the city.
High rates of hardship are reported in the south and west sides of the city where there are higher concentrations of people of color and higher instances of domestic violence reported.
The figure below illustrates the relative frequency of specific search terms over the course of the period that we were curious about.
We ran a chi-square goodness of fit test because we were comparing counts of crimes from the three different periods (p-val < .001).
$H_0$: there is no significant difference in the amount of total domestic violence crimes reported during lockdown when compared to the periods immediately preceding and immediately following lockdown.
$H_A$: There is a significant increase in the amount of domestic violence crimes reported to police during lockdown when compared to the periods immediately preceding and immediately following lockdown. Pre-lockdown and post-lockdown time periods do not show a significant difference in reported domestic violence crimes.
Finding: There is significant variation between the three periods (p < 0.001); however, the finding was in the opposite direction than hypothesized, where lockdown crime counts were the lowest compared to pre- and post-lockdown crime counts.
We were curious how hardship and White race % per community area were related and found that there was a nearly perfect negative correlation, where neighborhoods that are predominately White have low hardship levels and vice versa (p-val < .001).
We ran non-parametric (because of variable standard deviations) analyses of variance to test whether Google search volumes for particular keywords varied across three equal periods.
$H_0$: there is no significant difference in the amount of key terms searched related to domestic violence reported during lockdown when compared to the periods immediately preceding and immediately following lockdown.
$H_A$: There is a significant increase in the amount of amount of key terms searched related to domestic violence during lockdown when compared to the periods immediately preceding and immediately following lockdown. Pre-lockdown and post-lockdown time periods do not show a significant difference in reported key term searches.
Finding: There were no statistically significant differences in search volume across the three periods (all p-vals > .05)
The impact that COVID has had on domestic violence was highly time-dependent on time period examined. While we expected there to be an increase in domestic violence crimes reported during lockdown, there was actually a decrease in crimes reported. The increase in crimes reported did come during the post-lockdown period.
Google Trends API data showed no significant increase in any search terms related to domestic violence when comparing pre-lockdown, lockdown, and post-lockdown periods.
Maps based on data from the Chicago Health Atlas showed that higher rates of domestic violence crime were shown in areas in communities with higher hardship indices and higher volumes of people of color.
The expected trend of increased domestic violence crime reports did occur, but in a time frame not covered by our Alternate Hypothesis. Spending so much time at home during lockdown may still have made domestic violence victims more likely to seek help, but at a later time. It could be hypothesized that victims may not have wanted to have contact with more people than necessary (e.g. police) due to trying to avoid contracting COVID. Additionally, many services which may have served domestic violence victims may have been operating at a lesser capacity, and a victim seeking services could not, or would not, physically go to another location for help.
The use of the City of Chicago crime data set is limited to the amount of crimes reported- if fear of interacting with the police is a factor, particularly among communities of color, many more instances of domestic violence may go unreported. In addition, the offending family member may have threatened to take the victim's life, often causing the victim to keep violent occurrences to themself.
While we obtained one statistic from the Chicago Health Atlas data not directly related to rates of domestic violence crimes, there is visual evidence, via side-by-side comparison, that the most domestic violence crimes are occurring in areas with a high Hardship Index. Finding a common variable to merge this dataset with the City of Chicago crime data, and therefore run statistical analysis, would be one possible area of future study.
Data from the Google Trends API did not fall in line with the City of Chicago crime data, which was surprising because of the increase of dependence on the Internet to carry out daily life during the pandemic. However, domestic violence victims most likely knew there was a problem in their relationships pre-lockdown and may have already been looking into options for help.